9. Subagents

Notes

Subagents Claude can delegate tasks to subagents that break them down and run component tasks in parallel, improving your context management. Each subagent operates in its own isolated context window.

How It Works A subagent runs in parallel with its own context window, handles exploration or heavy lifting, and once finished, summarizes its findings and returns that summary back to Claude. The result: you get the answer you were looking for, without the entire journey cluttering your main context.

Creating Your Own Subagent Subagents are defined in Markdown files with YAML frontmatter. Run:

/agents

Then select "Create new agent." You'll walk through steps including choosing the scope, defining its purpose, selecting the tools it has access to, and picking a color. You also define when Claude should call the subagent automatically based on your prompts.

Further Customization

Persistent memory — lets your subagent retain memory across conversations
Preload skills into subagents — add the skill key and list skills by name. Note that unlike skills in your main conversation, the entire skill is loaded into context

Recap: With subagents, you can run an agent in the background to handle heavy lifting and return just the answer to your main context window. See also: Introduction to Subagents course.